Reporters’ Memo: Data Show Trade Had Greater Impact Than Automation on Manufacturing Job Loss

The Ohio Democratic debate launched the latest wave of stories wrongly telling people that automation – not trade – caused the loss of millions of U.S. manufacturing jobs in recent years. Jobs are lost for many reasons. But the “automation rather than trade story” is not supported by the data. Indeed, researchers increasingly are concluding that trade was the predominant cause of  manufacturing job loss in the past two decades. That bad news has a silver lining: Trade policy, unlike technological change, is subject to democratic decision-making. New trade policies can achieve improved outcomes.

Investment in automation slowed during the post-2000 period of mass manufacturing job loss, but during that period, the U.S. trade deficit exploded. Researchers have found that job displacement from technology is at its lowest level in decades.

  • Nearly five million U.S. manufacturing jobs have been lost since 2000 and one in five U.S. factories have closed.1
  • Yet, business investment rates in new equipment (eg. robotics) dropped from 5.1 percent during 1995-2002 to 1.7 percent between 2007-2016. The growth in business investment in IT equipment and software between 2007 and 2016 was less than one-third the rate between 1995 and 2002.2
  • Productivity growth (i.e., automation requiring fewer workers to produce each unit of output) actually dropped during the period of greatest manufacturing job loss – from 4.4% during the 1995-2000 IT-driven productivity boom to 2.2%3 between 2000 and 2018. 4 Workers were not replaced by robots, but lost jobs to imports and outsourcing. The extraordinary plunge in manufacturing jobs was due to the China-trade-related explosion of our trade deficit. It peaked at just under 6% of GDP ($1.2 trillion in today’s dollars in 2005 and 2006.) The Congressional Budget Office estimates that annual productivity growth over the next decade will be at 1.8%,  about the same as current levels and nowhere near the 1995 to 2000 rate from before the mass manufacturing job loss.5

Claims that automation caused manufacturing jobs loss are premised on misinterpretation of the data. The value of what is being produced in the U.S. manufacturing sector has grown even as millions of manufacturing jobs were lost. The popular view is that each manufacturing worker is producing more because factories were automated.6 Labor economist Susan Houseman at the Upjohn Institute showed that this story is based on the mistaken assumption that productivity growth reflects the rise of automation.7 

  • The growth in U.S. manufacturing output comes mainly from just one sector: computers and electronics. Overall manufacturing output today is only 8% higher than in the 1990s and remains lower than before the Great Recession. Thus, the premise of the story we’re being told about fewer manufacturing workers making more goods is wrong for most of manufacturing.
  • Much of the productivity growth in computers and electronics has little to do with automation, but rather reflects how government statisticians account for increases in product quality. Statisticians adjust so-called “price deflators” to account for the fact that computers are much faster than they used to be or have more memory. Thus, increases in real manufacturing output are being driven not by workers producing more computers, but by the production of better computers. Houseman explains how this distorts the headline indicators: “[T]he rapid productivity growth in the computer industry – and by extension, the strong productivity growth in manufacturing – largely reflects improvements in high-tech products, not automation.”8
  • That means that while measured manufacturing productivity may be increasing, U.S. real manufacturing output overall is actually growing slowly, not fast. According to Houseman and others, the main reason is that U.S. consumers are buying more goods produced overseas. As multinational companies take advantage of the ability to pay workers less and operate under lax environment regulations, she notes that the “locus of production” of many products has been shifting to Asia.9 If low-productivity, labor-intensive tasks are done abroad, then as a consequence, what’s left behind in the United States is measured to be more productive.10

Manufacturing more with fewer workers is not a new trend. What changed was growth in the trade deficit. We have had forms of automation in manufacturing for a century. Productivity growth has always meant that fewer manufacturing workers could produce the same amount of stuff. In the decades of the 1950s and 1960s, when productivity growth was far more rapid than it has been recent decades, it was associated with rising wages and low unemployment. Productivity growth is not new and is usually good for workers. What was new in the years from 2000 to 2007, when we lost more than 3 million manufacturing jobs, was the explosion in the trade deficit related to Congress’ 2000 approval of China’s entry into the World Trade Organization.

chart-of-manufacturing-employment-in-the-united-states

  • While manufacturing employment was declining as a share of total employment over this whole period, there was little change in the actual number of manufacturing jobs until 2000. (In 1970, the United States had 17.3 million manufacturing jobs; in 2000, 17.2 million manufacturing jobs.) There was enormous growth in manufacturing productivity over this period, yet we had very little change in total employment. Employment then plummeted from 2000 to 2007 (before the financial crash) as a result of an exploding trade deficit that peaked at almost 6 percent of GDP in 2005 and 2006.

Some of the studies demonstrating that trade, not automation, accounts for much of U.S. manufacturing job loss:

  • “Understanding the Decline of U.S. Manufacturing Employment,” Susan Houseman, Upjohn Institute for Employment Research Working Paper, January 2018: The findings of this paper are summarized above.11
  • “False Alarmism: Technological Disruption and the U.S. Labor Market, 1850-2015,” Robert Atkinson and John Wu, Information Technology & Innovation Foundation, May 2017: The researchers used U.S. Census Bureau data to review U.S. occupational trends from 1850 to 2010 and 2010 to 2015 to seek an empirical basis for claims of an acceleration in displacement of workers from workplace automation. They found that occupational churn – the absolute value sum of jobs added in growing occupations and jobs lost in declining occupations – is at the lowest level since 1850. And, in the past 15 years, the rate of losses are 50% of those in the 1960s, 1970s or 1990s. They also found that for the 2010-2015 period, about six technology-related jobs were created for every 10 lost, the lowest ratio of jobs-lost-to-jobs-gained for any period since 1950-1960.12
  • “Robots and Jobs: Evidence from U.S. Labor Markets,” Daron Acemoglu and Pascual Restrepo, National Bureau of Economic Research Working Paper 23285, March 2017: The research found only small negative employment effects from adoption of industrial robots. The authors acknowledged that U.S. trade with China from 1990 to 2007 displaced three times as many jobs as robots did.13
  • “Robots at Work,” Georg Graetz and Guy Michaels, Centre for Economic Performance Discussion Paper 1335, March 2015: A 2016 study within European countries found industry-level adoption off industrial robots raised worker wages and had no effect on employment.14
  • “The Decline of the U.S. Labor Share,” Michael W. L. Elsby, Bart Hobijn and Aysegul Sahin, The Brookings Institution, Brookings Papers on Economic Activity, Fall 2013: Economists at the Federal Reserve and University of Edinburgh identified why U.S. workers’ share of national income has been steadily declining over the past couple of decades. The authors find “limited support” for the theory that technological change primarily explains middle-class workers’ diminishing slice of the economic pie. Instead, they conclude, “[O]ur analysis identifies offshoring of the labor-intensive component of the U.S. supply chain as a leading potential explanation of the decline in the U.S. labor share over the past 25 years.”15 Indeed, their findings “suggest that increases in the import exposure of U.S. businesses can account for 3.3 percentage points of the 3.9 percentage point decline in the U.S. payroll share over the past quarter century.”16That is, import competition from 1993 through 2010 explains 85% of the observed decline in U.S. workers’ share of national income – a result that the economists find “striking,” leading them to suggest that if the trade status quo continues, “the labor share will continue to decline.”17
  • “Untangling Trade and Technology: Evidence from Local Labor Markets,” David H. Autor, David Dorn and Gordon H. Hanson, National Bureau of Economic Research Working Paper 18938, April 2013: This study on the U.S. job impacts of both technology and trade found “no net employment decline” from technological change from 1990 to 2007 while finding a strong correlation between increasing import competition from China and “significant falls in employment, particularly in manufacturing and among non-college workers.”18

Excerpts from Economist Dean Baker:
Trade Denialism Continues: Trade Really Did Kill Manufacturing Jobs,” Truthout

There have been a flood of opinion pieces and news stories in recent weeks wrongly telling people that it was not trade that led to the loss of manufacturing jobs in recent years, but rather automation. This means that all of those people who are worried about trade deficits costing jobs are simply being silly. The promulgators of the automation story want everyone to stop talking about trade and instead focus on education, technology or whatever other item they can throw out as a distraction…

The basic story on automation, trade and jobs is fairly straightforward. “Automation” is also known as “productivity growth,” and it is not new. We have been seeing gains in productivity in manufacturing ever since we started manufacturing things. Productivity gains mean that we can produce more output with the same amount of work. Before the trade deficit exploded in the last decade, increases in productivity were largely offset by increases in output, making it so the total jobs in manufacturing did not change much.

Imagine that productivity increased by 20 percent over the course of a decade, roughly its average rate of growth. If manufacturing output also increases by 20 percent, then we have the same number of jobs at the end of the decade as at the beginning. This is pretty much what happened before the trade deficit exploded.

This is easy to see in the data. In December of 1970 the US had 17.3 million manufacturing jobs. Thirty years later, in December of 2000, it had 17.2 million manufacturing jobs. We had enormous growth in manufacturing productivity over this period, yet we had very little change in total employment.

To be clear, manufacturing did decline as a share of total employment. Total employment nearly doubled from 1970 to 2000, which means that the share of manufacturing employment in total employment fell by almost half. People were increasingly spending their money on services rather than manufactured goods.

However what we saw in the years after 2000 was qualitatively different. The number of manufacturing jobs fell by 3.4 million, more than 20 percent, between December 2000 and December of 2007. Note that this is before the collapse of the housing bubbled caused the recession. Manufacturing employment dropped by an additional 2.3 million in the recession, although it has since regained roughly half of these jobs.

The extraordinary plunge in manufacturing jobs in the years 2000 to 2007 was due to the explosion of the trade deficit, which peaked at just under 6 percent of GDP ($1.2 trillion in today’s economy) in 2005 and 2006. This was first and foremost due to the growth of imports from China during these years, although we ran large trade deficits with other countries as well.

There really is very little ambiguity in this story. Does anyone believe that if we had balanced trade it wouldn’t mean more manufacturing jobs? Do they think we could produce another $1.2 trillion in manufacturing output without employing any workers?


1. U.S. Bureau of Labor Statistics, Current Employment Statistics survey, series ID CES3000000001, manufacturing

industry, 2019. Available at: https://data.bls.gov/timeseries/ces3000000001. “Quarterly Census of Employment and Wages,” County High Level Excel Files, manufacturing, number of establishments, 2018. Available at http://www.bls.gov/cew/datatoc.htm.

2. Lawrence Mishel and Josh Bivens, “The Zombie Robot Argument Lurches On,” Economic Policy Institute Report, May 24, 2017, at 9. Available at: https://www.epi.org/publication/the-zombie-robot-argument-lurches-on-there-is-no-evidence-that-automation-leads-to-joblessness-or-inequality/.

3. Aaron Cobet and Gregory Wilson, “International Comparisons: Comparing 50 Years of Labor Productivity in U.s. and Foreign Manufacturing,” U.S. Bureau of Labor Statistics Report, June 2002, at 61. Available at: https://www.bls.gov/opub/mlr/2002/06/art4full.pdf.

4. U.S. Bureau of Labor Statistics, Productivity Change in the manufacturing Sector, 1987-2018. Available at: https://www.bls.gov/lpc/special_requests
/mfgbardata.txt
.

5. Congressional Budget Office, “The Budget and Economic Outlook: 2019 to 2029,” Available at: https://www.cbo.gov/system/files
?file=2019-01/54918-Outlook.pdf
.

6. Binyamin Appelbaum, “Why are Politicians So Obsessed with Manufacturing,” The New York Times, Oct. 4, 2016. Available at: https://www.nytimes.com/2016/10/09/magazine/why-are-politicians-so-obsessed-with-manufacturing.html. Appelbaum writes: “There can be no revival of American manufacturing, because there has been no collapse. Because of automation, there are far fewer jobs in factories.”

7. Susan Houseman, “Understanding the Decline of U.S. Manufacturing Employment,” Upjohn Institute for Employment Research Working Paper, Jan. 2018. Available at: http://www.upjohn.org/mfg-decline.pdf.

8. Susan Houseman, “The Decline of U.S. Manufacturing Employment – Automation and Trade,” W.E. Upjohn Institute, April 2018. Available at: https://research.upjohn.org/cgi/viewcontent.
cgi?referer=&httpsredir=1&article=1268&
context=empl_research
. This is a condensed version of Houseman’s Jan. 2018 paper.

9. Susan Houseman, Timothy Bartick and Timothy Sturgeon, “Measuring Manufacturing; How the Computer and Semiconductor Industries Affect the Numbers and Perceptions,” in Susan Housem and Michael Mandel, eds., Measuring Globalization: Better Trade Statistics for Better Policy – Volume I. Biases to Price, Output, and Productivity Statistics from Trade, (Kalamazoo: W.E. Upjohn Institute for Employment Research, 2015).

10. Thomas Holmes, “The Case of the Disappearing Large-Employer Manufacturing Plants: Not Much of a Mystery After All,” Federal Reserve Bank of Minneapolis, Sept. 27, 2011. Available at: https://www.minneapolisfed.org/publications/the-region/the-case-of-the-disappearing-largeemployer-manufacturing-plants-not-much-of-a-mystery-after-all.

11. Susan Houseman, “Understanding the Decline of U.S. Manufacturing Employment,” Upjohn Institute for Employment Research Working Paper, Jan. 2018. Available at: http://www.upjohn.org/mfg-decline.pdf.

12. Robert Atkinson and John Wu, “False Alarmism: Technological Disruption and the U.S. Labor Market, 1850-2015,” Information Technology & Innovation Foundation Report, May 2017. Available at: http://www2.itif.org/2017-false-alarmism-technological-disruption.pdf?_ga=2.95546201.47652356.1549579395-1802087980.1549579395.

13. Daron Acemoglu and Pascual Restrepo, “Robots and Jobs: Evidence from U.S. Labor Markets,” National Bureau of Economic Research Working Paper 23285, March 2017. Available at: https://www.nber.org/papers/w23285.pdf.

14. Georg Graetz and Guy Michaels, “Robots at Work,” Centre for Economic Performance Discussion Paper 1335, March 2015. Available at: http://cep.lse.ac.uk/pubs/download/dp1335.pdf.

15. Michael W. L. Elsby, Bart Hobijn and Aysegul Sahin, “The Decline of the U.S. Labor Share,” Brookings Papers on Economic Activity, Fall 2013, at 1. Available at: http://www.brookings.edu/~/media/Projects
/BPEA/Fall%202013/2013b_elsby_labor_share.pdf
.

16. Id, at 43.

17. Id, at 4 and 47.

18. David H. Autor, David Dorn and Gordon H. Hanson, “Untangling Trade and Technology: Evidence from Local Labor Markets,” National Bureau of Economic Research, Working Paper 18938, April 2013, at Abstract. Available at: https://www.nber.org/papers/w18938.pdf.